[GH-ISSUE #12828] there is no way to control model download concurrency (ollama pull) #70561

Closed
opened 2026-05-04 21:59:37 -05:00 by GiteaMirror · 1 comment
Owner

Originally created by @FlorinAndrei on GitHub (Oct 29, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/12828

What is the issue?

Run ollama pull <some-model> on Linux. Examine the count of parallel download channels that are active at that time. It's a large number in my tests, perhaps something like a dozen. There are multiple Ollama threads, and multiple active TCP connections, all generating traffic.

This is helpful when the Ollama model repo is slow. It is very unhelpful when the repo is fast enough but your connection is not fast - then your internet link gets flooded with traffic, all your internet traffic slows down, and you get complaints from people sharing the link with you.

Seems like a small thing, but the impact is large.

Please allow some way to control download parallelism. Even something as simple as a binary switch (turn on or off the download multithreading) would be very helpful. I could then turn download concurrency off and let the models take their own sweet time to download, which is what I like to do anyway - but without nuking my internet. Thanks!

Relevant log output


OS

Linux

GPU

Nvidia

CPU

AMD

Ollama version

0.12.6

Originally created by @FlorinAndrei on GitHub (Oct 29, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/12828 ### What is the issue? Run `ollama pull <some-model>` on Linux. Examine the count of parallel download channels that are active at that time. It's a large number in my tests, perhaps something like a dozen. There are multiple Ollama threads, and multiple active TCP connections, all generating traffic. This is helpful when the Ollama model repo is slow. It is very unhelpful when the repo is fast enough but your connection is not fast - then your internet link gets flooded with traffic, all your internet traffic slows down, and you get complaints from people sharing the link with you. Seems like a small thing, but the impact is large. Please allow some way to control download parallelism. Even something as simple as a binary switch (turn on or off the download multithreading) would be very helpful. I could then turn download concurrency off and let the models take their own sweet time to download, which is what I like to do anyway - but without nuking my internet. Thanks! ### Relevant log output ```shell ``` ### OS Linux ### GPU Nvidia ### CPU AMD ### Ollama version 0.12.6
GiteaMirror added the bug label 2026-05-04 21:59:37 -05:00
Author
Owner

@rick-github commented on GitHub (Oct 29, 2025):

https://github.com/ollama/ollama/issues/10331

<!-- gh-comment-id:3463074151 --> @rick-github commented on GitHub (Oct 29, 2025): https://github.com/ollama/ollama/issues/10331
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#70561